The Risk of AI Optimisation: Becoming a Source, Not a Destination
- Thom Hayes
- Apr 30
- 7 min read
This is the question that keeps nagging at me with AI optimisation: what happens if a website becomes so easy to extract from, summarise and reuse that people barely have a reason to visit it?
That may be too simplistic, but I think there's some thought needed here.
A lot of the advice around AI visibility makes sense on paper. Clearer structure. Cleaner answers. Better organisation. More schema. More FAQs. Stronger service pages. Better supporting content. All sensible. But if all of that makes the answer easier to deliver inside the AI interface itself, what's left to pull the person through to the website?
That is the phrase I keep landing on:
A site can become a source without becoming a destination.
I don't think the answer is to become vague or to hold useful information back.
But I do think there is something worth paying attention to here. The more we shape pages to be legible to machines, the more important it becomes not to flatten out the parts that make them persuasive, distinctive and worth visiting for an actual person.
AI readability is useful, but it isn't the final test
The point isn't just to be understood by AI - A website should be easy to understand, full stop.
That's true for Google. It's true for AI systems. And it's true for human visitors too.
So this isn't an argument against clarity. It's not an argument against structure. It's not an argument against cleaner headings, stronger service pages or better organised content. Most of that is useful, and in many cases long overdue.
The thing I'm less sure about is what happens when readability becomes the whole goal.
The final test shouldn't be whether AI can understand the page, but whether a real person still wants to act on it.
A page can become very easy to summarise while losing some of the qualities that make it convincing. The tone can flatten out. The rhythm can become mechanical. The content can start to feel more like an answer bank than a page written for a person.
That may help with extraction, but it doesn't necessarily help with persuasion.
A website can be useful without being visited
That feels like the real risk underneath all this.
If a page is structured in a way that gives away everything in a neat, clinical, self-contained answer, it may become highly useful to AI systems without giving a person much reason to click through.
That doesn't mean the visibility has no value. It may still help with awareness. It may still help with credibility. It may still help your business surface in relevant conversations.
But the question is whether that's enough on its own.
If the site is doing a good job of powering the answer, but not much of a job of pulling people in, then part of the value has shifted away from the site itself. The content is still working, but perhaps not in the way most businesses instinctively want it to.
That's where the source versus destination point starts to matter.
A source is where the answer comes from. A destination is where the person goes when they want the fuller picture, the nuance, the examples, the proof, the visuals, the process, or the confidence to take the next step.
If a website only succeeds at being the first of those, there's a limit to how much practical value that visibility creates.
This probably matters more in crowded markets
The risk isn't evenly spread.
If a business offers something genuinely distinctive, niche or hard to flatten into a generic category, the risk may be lower. In those cases, AI visibility may still help clarify the offer and move someone closer to the business.
The issue is weighted more towards saturated markets, where lots of providers offer broadly similar services and answer broadly similar questions.
That's where AI summaries may be more likely to compress the landscape, reduce the need to explore individual websites, and make it harder for any one business to become much more than a source.
The answer isn't to become less useful
I don't think the solution is to hold the answer back.
That would be the wrong response. A weak, evasive page isn't more effective just because it withholds detail. If anything, that would make it less useful for everybody.
The better question is how to make a page genuinely helpful while still leaving room for the richer layer that makes a visit worthwhile.
That's a different challenge. It means answering the question clearly, but not reducing the whole page to one dead-end summary. It means being useful without becoming disposable. It means giving the essentials away while still making the page feel like the place where the answer becomes more usable.
That richer layer might be:
More depth
More nuance
Better explanation
Practical trade-offs
Visuals
Process
Proof
Examples
Comparisons
Next Steps
That's where a page starts to become more than something quotable.
A strong page probably works in layers
A useful first layer should lead into a stronger second one.
The first layer is what helps the page surface. It's the clear answer, the understandable structure, the obvious relevance. It's the part that helps a machine, and usually a skim-reading human too, understand what the page is about quickly.
The second layer is what makes the page worth spending time with.
That's often where the real value sits. Not because the page becomes long for the sake of it, but because it becomes more usable, more trustworthy and more human:
A service page might explain what the service is in a few clear lines near the top. That's useful, but the rest of the page still needs to help the reader decide whether it feels right, how it works, what makes it different, and what to do next.
A blog post might give the core takeaway early. That's useful too, but the rest of the article still needs to add context, judgment and a stronger way of thinking about the issue, rather than simply repeating the top-line summary in slightly different words.
That layered approach feels like the healthier balance. It makes the page easier to understand without making it feel redundant.
Traditional SEO still matters because people still search like people
Not every meaningful visit starts inside an AI answer.
This is easy to forget once the conversation gets too AI-heavy.
People still search in ordinary ways. They still click through from search results. They still compare options. They still decide quickly whether something feels clear, credible and relevant. That means the older fundamentals still matter.
Strong title tags still matter. Clear H1s still matter. Proper service pages still matter. Search-led blog topics still matter. Internal links still matter. Content that sounds like it was written by a person still matters.
Traditional SEO was never just about pleasing Google.
At its best, it was about making websites clearer, more useful and easier to find. That still feels like a big part of the answer now.
If AI optimisation pulls a site too far away from that, the result may be a page that is easier for machines to digest but weaker for the people who actually make decisions.
A website still has to convert like a website
The person who lands there still needs a reason to trust it.
This, to me, is where the human side becomes impossible to ignore. The end goal isn't an AI system understanding the page. The end goal is a real person deciding to do something.
That could mean making contact. Booking a call. Sending an enquiry. Reading more. Comparing services. Looking at examples. Trusting the business enough to move forward.
If the page feels too robotic, too generic or too flattened, that becomes harder.
A website can be easy for AI to summarise and still do a poor job of converting a real person.
A convincing page still needs tone. It still needs clarity. It still needs structure. But it also needs a sense that there is a real person or real business behind it, and that the page has been shaped with some care rather than reduced to a machine-friendly splurge.
The human part isn't something to preserve out of sentiment. It's part of what makes the site work.
There may be a second upside to stronger AI-ready content
The click isn't the only possible next step.
This is the other side of the thought. If a website is optimised for AI systems to understand properly, that may still create useful momentum even if the first click never happens.
Someone may not visit the site immediately, but they might continue the conversation.
They may ask more about the business. They may ask what the company offers, who it is for, how it differs from others, what the process is, or whether it sounds like the right fit.
That matters, because if your website does not give AI enough to work with at that point, the conversation may simply move elsewhere. If the obvious follow-up questions are answered more clearly on a competitor’s site, there is every chance that competitor becomes the more useful source for the next stage of this customer's journey.
So the answer isn’t to optimise less for AI. In some ways, it may be to optimise more thoughtfully. Informative service pages, useful supporting content, sensible FAQs and structured signals such as schema can all help build a fuller picture. The point isn’t just to answer the first question. It’s to stay useful as the questions become more specific.
The goal may not always be to earn the click straight away.
Sometimes it may be to earn the next question, and still be the best place for the answer when that question comes.
That still makes stronger content worthwhile. But the website also needs to be ready for the moment a person does finally click through. If that click arrives later, the site still has to feel human, credible and worth acting on.
The human / AI balance
The best websites are understandable to machines and persuasive to people.
That feels like the healthier target. Not content that hides the answer. Not content that gives everything away in one neat disposable summary. And not content that sounds like it was assembled purely for extraction.
The stronger route is probably to create pages that are:
Clear enough to be interpreted
Structured enough to be surfaced
Useful enough to answer the question
Human enough to build trust
Distinct enough to be remembered
Strong enough to convert when someone clicks through
Easy enough for machines to understand. Strong enough for people to trust.
That's the real balance as I see it.
A better way to think about AI optimisation
The goal isn't just to be quotable. It's to remain worth visiting.
Yes, it matters if AI systems can understand your site, your pages can surface clearly, and your content is structured well enough to be useful in those environments. But if all of that happens at the expense of the page feeling human, persuasive and worth acting on, then something important has been lost.
The aim isn't to reject AI optimisation, but to avoid doing it in a way that forgets or trims off the parts that make a website work in the first place.
Because a site can become a source without becoming a destination. And if that happens, the visibility may still be there, but the visit, the trust and the action may be much harder to win.